[Intel-gfx] [PATCH] drm/i915: Inspect subunit states on hangcheck

Arun Siluvery arun.siluvery at linux.intel.com
Tue Dec 1 04:56:55 PST 2015


On 01/12/2015 12:17, Mika Kuoppala wrote:
> If head seems stuck and engine in question is rcs,
> inspect subunit state transitions before deciding that
> this really is a hang instead of limited progress.
>
> References: https://bugs.freedesktop.org/show_bug.cgi?id=93029
> Cc: Chris Wilson <chris at chris-wilson.co.uk>
> Cc: Dave Gordon <david.s.gordon at intel.com>
> Cc: Daniel Vetter <daniel at ffwll.ch>
> Signed-off-by: Mika Kuoppala <mika.kuoppala at intel.com>
> ---
>   drivers/gpu/drm/i915/i915_irq.c         | 49 +++++++++++++++++++++++++++++----
>   drivers/gpu/drm/i915/intel_ringbuffer.h |  1 +
>   2 files changed, 45 insertions(+), 5 deletions(-)
>
> diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c
> index e88d692..e6ae54f 100644
> --- a/drivers/gpu/drm/i915/i915_irq.c
> +++ b/drivers/gpu/drm/i915/i915_irq.c
> @@ -2913,13 +2913,31 @@ static void semaphore_clear_deadlocks(struct drm_i915_private *dev_priv)
>   		ring->hangcheck.deadlock = 0;
>   }
>
> -static enum intel_ring_hangcheck_action
> -ring_stuck(struct intel_engine_cs *ring, u64 acthd)
> +static bool subunits_stuck(struct intel_engine_cs *ring)
>   {
> -	struct drm_device *dev = ring->dev;
> -	struct drm_i915_private *dev_priv = dev->dev_private;
> -	u32 tmp;
> +	int i;
> +	u32 instdone[I915_NUM_INSTDONE_REG];
> +	bool stuck;
> +
> +	if (ring->id != RCS)
> +		return true;
> +
> +	i915_get_extra_instdone(ring->dev, instdone);
>
> +	stuck = true;
> +	for (i = 0; i < I915_NUM_INSTDONE_REG; i++) {
> +		if (instdone[i] != ring->hangcheck.instdone[i])
> +			stuck = false;

This may not be completely reliable. Tomas Elf in his TDR tests observed 
that instdone kept changing even when CS is hung and in a stable state.

regards
Arun

> +
> +		ring->hangcheck.instdone[i] = instdone[i];
> +	}
> +
> +	return stuck;
> +}
> +
> +static enum intel_ring_hangcheck_action
> +head_stuck(struct intel_engine_cs *ring, u64 acthd)
> +{
>   	if (acthd != ring->hangcheck.acthd) {
>   		if (acthd > ring->hangcheck.max_acthd) {
>   			ring->hangcheck.max_acthd = acthd;
> @@ -2929,6 +2947,24 @@ ring_stuck(struct intel_engine_cs *ring, u64 acthd)
>   		return HANGCHECK_ACTIVE_LOOP;
>   	}
>
> +	if (!subunits_stuck(ring))
> +		return HANGCHECK_ACTIVE_LOOP;
> +
> +	return HANGCHECK_HUNG;
> +}
> +
> +static enum intel_ring_hangcheck_action
> +ring_stuck(struct intel_engine_cs *ring, u64 acthd)
> +{
> +	struct drm_device *dev = ring->dev;
> +	struct drm_i915_private *dev_priv = dev->dev_private;
> +	enum intel_ring_hangcheck_action ha;
> +	u32 tmp;
> +
> +	ha = head_stuck(ring, acthd);
> +	if (ha != HANGCHECK_HUNG)
> +		return ha;
> +
>   	if (IS_GEN2(dev))
>   		return HANGCHECK_HUNG;
>
> @@ -3064,6 +3100,9 @@ static void i915_hangcheck_elapsed(struct work_struct *work)
>   				ring->hangcheck.score--;
>
>   			ring->hangcheck.acthd = ring->hangcheck.max_acthd = 0;
> +
> +			memset(ring->hangcheck.instdone, 0,
> +			       sizeof(ring->hangcheck.instdone));
>   		}
>
>   		ring->hangcheck.seqno = seqno;
> diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h
> index 5d1eb20..b8fe92e 100644
> --- a/drivers/gpu/drm/i915/intel_ringbuffer.h
> +++ b/drivers/gpu/drm/i915/intel_ringbuffer.h
> @@ -93,6 +93,7 @@ struct intel_ring_hangcheck {
>   	int score;
>   	enum intel_ring_hangcheck_action action;
>   	int deadlock;
> +	u32 instdone[I915_NUM_INSTDONE_REG];
>   };
>
>   struct intel_ringbuffer {
>



More information about the Intel-gfx mailing list